BACKGROUND Clinical predictors associated with acute paraquat (PQ) poisoning have not been systematically studied. OBJECTIVE To identify independent predictors of death in patients with acute PQ poisoning. METHODS This is a retrospective study executed in the emergency department of a university hospital. One hundred three consecutive patients poisoned with PQ between January 1999 and December 2004 were enrolled. Urine PQ concentration, electrolyte and renal function, detailed history, and Acute Physiology and Chronic Health Evaluation II were extracted from medical records. The outcome measure was 30-day mortality. Multivariate analysis was done by Cox-proportional hazard regression model. Receiver operating characteristics area under the curve was calculated for selected predictors. RESULTS The crude 30-day mortality was 67.9% (70 of 103). Independent predictors of death were acute renal failure (hazard ratio, 3.53; 95% confidence interval, 1.97-6.32), hypokalemia (2.07, 1.21-3.51), hypothermia (2.91, 1.67-5.07), suicide (2.11, 1.04-4.29), and self-reported ingested dose (2.06, 1.38-3.06). The receiver operating characteristics area under the curve of serum potassium concentrations, maximal urine PQ concentrations, and Acute Physiology and Chronic Health Evaluation II scores were 0.75 (95% confidence interval, 0.60-0.81), 0.71 (0.66-0.84), and 0.80 (0.71-0.88), respectively. Under the cutoff value of 3.6 mEq/L, hypokalemia had a sensitivity of 75% and specificity of 54% in predicting mortality. CONCLUSION The identified risk factors may allow better identification of those at greater mortality risk. Future development of a tailored clinical scoring system incorporating the identified risk factors for acute PQ poisoning may be of great help.